ProtoGene.pl 57.2 KB
Newer Older
Sebastien Moretti's avatar
Sebastien Moretti committed
1
2
3
4
5
6
7
8
9
10
11
12
#!/usr/bin/env perl

########################################################
#                                                      #
# ProtoGene : Bona-Fide Back Translation of Protein    #
#             Multiple Sequence Alignments             #
#                                                      #
########################################################

use strict;
use warnings;
use diagnostics;
Sebastien Moretti's avatar
Sebastien Moretti committed
13
14
use Carp;

Sebastien Moretti's avatar
Sebastien Moretti committed
15
use File::Which qw(which);                 # Locate external executable programs in the PATH
16
17
use Time::localtime;                       # Use localtime+PID for a pseudo-uniq temp file name
use Getopt::Long;                          # Options specifications
Sebastien Moretti's avatar
Sebastien Moretti committed
18
use File::Copy qw(move);                   # Avoid external 'mv' command usage
19
use LWP::Simple;                           # To test gigablaster availability
Sebastien Moretti's avatar
Sebastien Moretti committed
20
21
22

use Mail::Send;                            # Send warnings and errors files by e-mail ==> only if the $userEMail variable is defined

23
24
use lib '/mnt/common/share/ProtoGene/';    # Local path for ProtoGene's own perl modules
###use lib '/mnt/local/lib/tcoffee_perl/';
Sebastien Moretti's avatar
Sebastien Moretti committed
25
use Exonerate;                             # Exonerate runner, parser, ...
26
use Views;                                 # Non-text outputs, e.g. HTML/CSS
27
#use CheckOutput;                         # Check output for cds consistancy with query
Sebastien Moretti's avatar
Sebastien Moretti committed
28
29
30



31
################## CONFIGURATION ##################
Sebastien Moretti's avatar
Sebastien Moretti committed
32
$ENV{'PATH'}        .= ':/mnt/local/bin/:./';                               # Additional path for executables
33
34

my $cachePath        = '/scratch/cluster/monthly/t_coffee/ProtoGene_Cache'; # Cache directory
35
$cachePath = '/Users/smoretti/Documents/ProtoGene/TTMMPP'; #FIXME
36
37
38
my $cacheStorageTime = 15;                                                  # Do not update sequences younger than X days

my $userEMail        = 'moretti.sebastien@gmail.com';                       # To receive e-mails with encountered problems; leave blank to inactive
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55

### BLAST parameters ###
my $blast_param = { 'evalue'    => 0.05,
                    'matrix'    => 'BLOSUM62',
                    'filter'    => 'Off',
                    'species'   => 'All_organisms',

                    'db1'       => 'refseq_protein', # NCBI RefSeq protein
                    'identity1' => 100,
                    'cover1'    => 100,
                    'db2'       => 'nr', # NCBI nr protein
                    'identity2' =>  95,
                    'cover2'    =>  95,

                    'nbesthits' => 10,
                    'core'      => 1,
                  };
56
###################################################
Sebastien Moretti's avatar
Sebastien Moretti committed
57
58


59
my $VERSION  = '4.0.0';
Sebastien Moretti's avatar
Sebastien Moretti committed
60

Sebastien Moretti's avatar
Sebastien Moretti committed
61
my $webblast_exe  = 'webblast.pl';
62
my $blast_exe     = 'blastall';            # Or wu-blastall for Wu-BLAST; for local blast usage
Sebastien Moretti's avatar
Sebastien Moretti committed
63
my $exonerate_exe = 'exonerate-1.0';       # Exonerate 1.0 because current parser only works with this version
64
65
66
67
68


################## Option management
my ($msa, $revtrans, $pep, $hideBOJ, $run_name, $template, $lim, $cache) = ('', 0, 0, 0, '', '', 0, 'update');
my ($debug, $tmp)                                                        = (0, 0);
69
my ($db, $species, $local, $giga)                                        = ($blast_param->{'db1'}, $blast_param->{'species'}, 0, 0);
Sebastien Moretti's avatar
Sebastien Moretti committed
70
71
72
73
74
75
76
77
78
79
80
my %opts = ('msa|in=s'        => \$msa,        # Input sequences
            'revtrans'        => \$revtrans,   # Use to reverse-translate sequences with no match
            'pep'             => \$pep,        # Add the original peptide query beneath the related CDS seq
            'hideBOJ'         => \$hideBOJ,    # Hide BOJ output
            'run_name=s'      => \$run_name,   # Use another name, instead of input seq name, for result files
            'template=s'      => \$template,   # Use a template file
            'lim=i'           => \$lim,        # Limit number of input query sequences
            'cache=s'         => \$cache,      # Cache behavior

            'orgm|species=s'  => \$species,    # Organism(s) to blast against
            'db|database=s'   => \$db,         # Database to blast against
81
            'local'           => \$local,      # Use to specify a local db query, defined in $local_db
Sebastien Moretti's avatar
Sebastien Moretti committed
82
83
84
85
86
87
88
89
90
91
            'giga'            => \$giga,       # Use GigaBlaster server

            'version|V'       => sub { print "\n\tPROTOGENE : Bona-Fide Back Translation of Protein Multiple Sequence Alignments\n\tversion   : $VERSION\n\n";
                                       exit 0;
                                 },            # Print version information
            'help|H|?'        => sub { system('perldoc', "$0")==0 || print {*STDERR} "\n\tCannot open the documentation with 'perldoc'\n\n";
                                       exit 0;
                                     },        # Print full help message
            'debug'           => \$debug,      # Verbose output
            'tmp'             => \$tmp,        # To keep traces of fake intermediate files like fake xml from NCBI, fake aln, ...
Sebastien Moretti's avatar
Sebastien Moretti committed
92
           );
Sebastien Moretti's avatar
Sebastien Moretti committed
93
94
my $test_option_values = Getopt::Long::GetOptions(%opts);

95
96

################## Short help message
97
if ( !$test_option_values || ($msa eq '' && $cache ne 'empty' && $cache ne 'old') ){
98
99
100
101
102
103
104
105
    print {*STDERR} "\n\tCannot open the MSA file in FASTA format
\tTry:  $0 --msa=path_of_the_fasta_msa_file [Options]

\tOptions: --orgm=All_organisms, Bacteria, Viruses, Vertebrata,
\t                Eukaryota, Mammalia, Primates, Homo_sapiens,
\t                Gallus_gallus, Bos_taurus, Escherichia_coli,
\t                Arabidopsis_thaliana, Mus_musculus,
\t                Drosophila_Melanogaster, ...
106
\t                default is '$blast_param->{'species'}'
107
\t         --db=nr, pdb, swissprot, refseq_protein
108
\t                default is '$blast_param->{'db1'}'
109
110
111
112
\t         --local to execute a local BLAST query with
\t                --db=path_for_a_local_db_blast_formated\n
\t         --template to provide your own nucleotidic sequences
\t                following the cds file format
113
\t         --revtrans reverse-translates sequences with no
114
115
\t                blast hit, in IUB (IUPAC) depiction code
\t                They are removed from the alignement by default
116
\t         --pep  adds the original peptide query beneath the
117
118
119
120
\t                back-translated sequence
\t         --cache=none, update, use, 'own_PATH_directory', old, empty
\t                to select the cache mode
\t                default is 'update'\n
121
122
123
\t         --debug   prints extra information when running
\t         --version prints version information
\t         --help    prints a full help message\n\n";
124
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
125
}
126
127


128
129
################## Check cache directory accessibility
checkCacheAccessibility($cachePath);
130

131
132
# Cache management
$cache = cacheManagement($cache, $cachePath, 0) || '.';
133

134
135
# Remove old files from the cache directory everytime ProtoGene is running
cacheManagement('old', $cachePath, 1);
136
137


138
139

################## Check external software in the PATH
140
checkExternalSoftware($webblast_exe, $exonerate_exe);
141
142
if ( $local ){
    checkExternalSoftware($blast_exe);
Sebastien Moretti's avatar
Sebastien Moretti committed
143
144
145
146
}



147
148
149
################## Check input files
# Open and check template file
my ($template_NT, $template_AA, $template_SEQ);
150
#TODO: template is fasta only now
151
152
($template_NT, $template_AA, $template_SEQ) = checkTemplate($template) if ( $template );

Sebastien Moretti's avatar
Sebastien Moretti committed
153

Sebastien Moretti's avatar
Sebastien Moretti committed
154
155
156
157
158
# Open and Check the input fasta file
my $originalMSA    = $msa;
my $converted      = 0; # Accept only fasta format
#TODO: Do a better file converter and identifier is msa is not fasta
#($msa, $converted) = isFastaFile($msa, $converted);
Sebastien Moretti's avatar
Sebastien Moretti committed
159

160
my $fasta_checker = -1;
161
my (@original_names, @original_seq); #Use 2 lists in parallel : original_names, seq
Sebastien Moretti's avatar
Sebastien Moretti committed
162
163
open( my $MSA, '<', "$msa" ) or die "\n\tCannot open your MSA file\n\n";
READ_MSA:
164
while(<$MSA>){
Sebastien Moretti's avatar
Sebastien Moretti committed
165
166
    next READ_MSA if ( $_ =~ /^#/ );

167
    if ( $_ =~ /^>/ ){
168
169
        $fasta_checker++;
        my $name = $_;
Sebastien Moretti's avatar
Sebastien Moretti committed
170
171
172
        $name    =~ s{\r\n}{}g; #Remove return lines from windows OS '^M'
        $name    =~ s{\r}{\n}g;
        $name    =~ s{^>[ \t]+}{>};
173
        chomp($name);
Sebastien Moretti's avatar
Sebastien Moretti committed
174
175

        push @original_names, $name;
Sebastien Moretti's avatar
Sebastien Moretti committed
176
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
177
178
179
180
181
182
183
184
185
186
    elsif ( $_ =~ /[A-Z\-\.]/i ){
        my $sequence = $_;
        $sequence    =~ s{\r\n}{}g;
        $sequence    =~ s{\r}{\n}g;
        $sequence    =~ s{\.}{-}g; #for msa with '.' as gap
        #TODO because exonerate cannot work with selenocystein
        $sequence    =~ s{[BJOUZ]}{X}ig; #U here for selenocystein
        $sequence    =~ s{[^A-Za-z\-\*\n]}{}g; #Remove all the non-gap or non-alphabetic characters from the seq
        chomp($sequence);

Sebastien Moretti's avatar
Sebastien Moretti committed
187
        #fasta sequence on 1 line
188
        if ( !exists($original_seq[$fasta_checker]) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
189
            push @original_seq, $sequence;
190
191
        }
        else {
Sebastien Moretti's avatar
Sebastien Moretti committed
192
            $original_seq[$fasta_checker] .= $sequence;
193
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
194
195
    }
}
196
close $MSA;
Sebastien Moretti's avatar
Sebastien Moretti committed
197
198

#Are there FASTA sequences ?
199
if ( $fasta_checker == -1 ){
Sebastien Moretti's avatar
Sebastien Moretti committed
200
    failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
201
202
    print {*STDERR} "\tThe MSA file does NOT seem to be a protein FASTA format
\tSee an example of FASTA format <a href='http://en.wikipedia.org/wiki/FASTA_format' target='_blank'>here</a>\n\n";
203
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
204
}
205
elsif ( $lim >0 && $fasta_checker > $lim ){
Sebastien Moretti's avatar
Sebastien Moretti committed
206
    failure();
207
    print {*STDERR} "\tThe FASTA file is too large, try with less than $lim sequences\n\tor split your file\n\n";
208
209
    exit(1);
}
210
elsif ( exists( $original_seq[0] ) && $original_seq[0] =~ /[acgtu]/i ){
211
    my $first_seq = $original_seq[0];
Sebastien Moretti's avatar
Sebastien Moretti committed
212
213
214
215
216
217
    #Check if sequences are amino acids and not nucleotides ONLY on the 1st sequence
    my $a         = ($first_seq =~ s{[aA]}{}g)               || 0;
    my $c         = ($first_seq =~ s{[cC]}{}g)               || 0;
    my $g         = ($first_seq =~ s{[gG]}{}g)               || 0;
    my $t         = ($first_seq =~ s{[tTuU]}{}g)             || 0;
    my $non       = ($first_seq =~ s{[^aAcCgGtTuUXxNn-]}{}g) || 0;
218
    if ( ($a+$c+$g+$t) >= (($a+$c+$g+$t+$non)*80/100) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
219
        failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
220
221
        print {*STDERR} "\tYour sequences seem already to be nucleotides
\tthe goal of this program is to turn AMINO ACID alignments into CDS nucleotide alignments\n\n";
222
223
        exit(1);
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
224
225
226
227
}
undef $fasta_checker;


Sebastien Moretti's avatar
Sebastien Moretti committed
228
# Check GigaBlaster status if used
229
if ( $giga==1 ){
230
    $local = 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
231
    $local = 2 if ( head('http://www.igs.cnrs-mrs.fr/public_g/remoteblast.cgi') );
232
233
}

Sebastien Moretti's avatar
Sebastien Moretti committed
234

Sebastien Moretti's avatar
Sebastien Moretti committed
235
# Temporary file extension
236
my $date = sprintf("%d-%02d-%02d_%02dh%02d", localtime->year() + 1900, localtime->mon(), localtime->mday(), localtime->hour, localtime->min);
Sebastien Moretti's avatar
Sebastien Moretti committed
237
238


239

Sebastien Moretti's avatar
Sebastien Moretti committed
240
#Start main program with version # of programs and list original queries
241
print "\n\t   Protogene\t$VERSION\t$date\n\n";
242

243
open(my $EXONERATEISHERE, "$exonerate_exe --version |");
244
my $IsExonerateHere = <$EXONERATEISHERE>;
245
close $EXONERATEISHERE;
Sebastien Moretti's avatar
Sebastien Moretti committed
246
247
print "\t   $IsExonerateHere\n";
undef $IsExonerateHere;
248
undef $VERSION;
249

Sebastien Moretti's avatar
Sebastien Moretti committed
250
for(my $m=0; $m<=$#original_names; $m++){
251
    printf("\t>%-3s  is  %s\n", $m, $original_names[$m]);
Sebastien Moretti's avatar
Sebastien Moretti committed
252
}
Sebastien Moretti's avatar
Sebastien Moretti committed
253
$date .= "_$$"; #Add PID to be more uniq
Sebastien Moretti's avatar
Sebastien Moretti committed
254
255
256
257



# run_name implementation for web server
258
$originalMSA = $run_name if ( $run_name ne '' );
Sebastien Moretti's avatar
Sebastien Moretti committed
259

260
261
unlink("$originalMSA.cds", "$originalMSA.cdsP", "$originalMSA.cdsP.html",
       "$originalMSA.out", "$originalMSA.boj",  "$originalMSA.log");
Sebastien Moretti's avatar
Sebastien Moretti committed
262
263


264
# Build the sequences, from the alignment, to perform blast search
265
open(my $LOG, '>', "$date.log") or die "\n\tCannot open LOG file\n\n";
266
EACH_SEQ:
267
268
269
270
271
272
for(my $r=0; $r<=$#original_names; $r++){
    my $fasta_header = $original_names[$r];
    my $short_name   = $original_names[$r];
    $fasta_header    =~ s{^>}{};               # Full FASTA header
    $short_name      =~ s{^([^\s\t]+).*$}{$1}; # Short name without description

Sebastien Moretti's avatar
Sebastien Moretti committed
273

Sebastien Moretti's avatar
Sebastien Moretti committed
274
    my @equivalent_blast_hits; #Multiple BLAST hits with the same highest e-value/score
275
276

    if ( exists($template_NT->{$short_name}) && $template_NT->{$short_name} ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
277
278
279
        # First try nucleotide template
        push @equivalent_blast_hits, 'My_own_seq';
        print "\n\n$r done\n\nNucleotide template for $r:\t$template_NT->{$short_name}\n\n*******************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
280
    }
281
    elsif ( exists($template_AA->{$short_name}) && $template_AA->{$short_name} ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
282
283
284
        # Second try amino acid/BLASTp template
        push @equivalent_blast_hits, $template_AA->{$short_name};
        print "\n\n$r done\n\nBlastP template for $r:\t$template_AA->{$short_name}\n\n*******************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
285
286
    }
    else{
287
    # else send them to webblast
Sebastien Moretti's avatar
Sebastien Moretti committed
288
289
290
291
292
293
294
        # Prepare sequence $r for BLAST
        open(my $READY2BLAST, '>', "$cache/$date.seq2blast.fas.$r");
        my $noGap = $original_seq[$r];
        $noGap    =~ s{\-}{}g; #remove gaps
        print {$READY2BLAST} ">$r\n$noGap\n";
        close $READY2BLAST;

295
296
297
298
299
        if ($r==0){
            # Show BLAST parameters only for the first BLAST run
            display_blast_param();
        }
        run_BLAST("$cache/$date.seq2blast.fas.$r", $local, $blast_param->{'db1'}, $blast_exe, "$cache/$date.blastp.$r");
Sebastien Moretti's avatar
Sebastien Moretti committed
300

301
302
303

        # Get the BLASTp hit(s)  acc number
        open(my $BLASTP, '<', "$cache/$date.blastp.$r");
304
        while(<$BLASTP>){
305
            push @equivalent_blast_hits, $1  if ( $_ =~ /^>\s*\d+@?\w?\w?\w?__([^\s\|\.]+).*$/ ); #Warning: double '_'
306
307
        }
        close $BLASTP;
308
        unlink("$cache/$date.blastp.$r", "$cache/$date.seq2blast.fas.$r") if ( $tmp == 0 || exists($equivalent_blast_hits[0]) );
Sebastien Moretti's avatar
Sebastien Moretti committed
309
        if ( !exists($equivalent_blast_hits[0]) ){
310
            print "\n$r ...\tNo blast result found above thresholds for $fasta_header\n\n";
311
            buildFailureOutputFiles($r, 'No_BLASTp_Result', 'Not_searched', '');
312
313
314
315
            undef $original_seq[$r];
            undef $original_names[$r];
            next EACH_SEQ;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
316
317
318
319
    }



320
    #When multiple equivalent blast hits  ==> use, and add, every hits
321
    my ($resultPOS, $resultBOJ) = ('', '');
322
    my @failureStatus = (''); # To always have $failureStatus[0] defined
323
    my @intronStatus  = ('');
324
    HIT_LINK:
Sebastien Moretti's avatar
Sebastien Moretti committed
325
326
    for(my $qq=0; $qq<=$#equivalent_blast_hits; $qq++){
        print "\n$fasta_header  -->  [$equivalent_blast_hits[$qq]]\n";
327
328
329
330


        my @nt_GIs;
        my $intronStep = 0;
331
332
333
        # Get Nt sequence from template
        if ( exists($template_NT->{$short_name}) && $template_NT->{$short_name} ne '' ){
            if ( $template_NT->{$short_name} !~ /^My_Seq$/i ){
334
                download_seq($cache, '', '', $template_NT->{$short_name});
335
336
337
338
339
340
341
342
                @nt_GIs = $template_NT->{$short_name};
            }
            else {
                open(my $MYSEQ, '>', "$cache/My_Seq-$$.fas");
                print {$MYSEQ} ">My_Seq-$$ for $short_name\n".$template_SEQ->{$short_name}."\n";
                close $MYSEQ;
                @nt_GIs = "My_Seq-$$";
            }
343
        }
344
        # Get Nt sequence through prot acc -> prot gi -> nt linked gi -> nt acc
345
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
346
            # Prot ACC -> PUID
Sebastien Moretti's avatar
Sebastien Moretti committed
347
            my $protGI = blastPAcc2PGI($equivalent_blast_hits[$qq]);
348
            if ( $protGI eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
349
350
                print "\tNo protein (prot GI) link found for $equivalent_blast_hits[$qq] in $fasta_header\n\n";
                @failureStatus = ('PUI_unavailable', $equivalent_blast_hits[$qq]);
351
                next HIT_LINK;
352
            }
Sebastien Moretti's avatar
Sebastien Moretti committed
353

Sebastien Moretti's avatar
Sebastien Moretti committed
354
            # PUID -> nt GIs & GeneID
Sebastien Moretti's avatar
Sebastien Moretti committed
355
            my ($ntGIs, $geneID) = protGI2NTGIs($protGI, $equivalent_blast_hits[$qq]);
356
            if ( $ntGIs eq '' && $geneID eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
357
358
                print "\tNo nucleotide (nt GIs) link found for $equivalent_blast_hits[$qq] in $fasta_header\n\n";
                @failureStatus = ('No_nt_link', $equivalent_blast_hits[$qq]);
359
                next HIT_LINK;
360
361
362
363
            }
            print " linked with [$ntGIs $geneID]\n";


Sebastien Moretti's avatar
Sebastien Moretti committed
364
            # Get transcript and contig seq
Sebastien Moretti's avatar
Sebastien Moretti committed
365
            if ( $ntGIs ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
366
                @nt_GIs = split(/,/, $ntGIs);
Sebastien Moretti's avatar
Sebastien Moretti committed
367
368
                downloadSeqFromGIs($cache, '', '', @nt_GIs);
            }
369

Sebastien Moretti's avatar
Sebastien Moretti committed
370
            # GeneID -> Chr
371
372
373
374
375
376
377
            if ( $geneID ne '' ){
                my @geneIDs = split(/,/, $geneID);
                while(<@geneIDs>){
                    my $geneID = $_;
                    my ($chr, $amont, $aval) = geneID2Chr($geneID, $equivalent_blast_hits[$qq]);
                    print "\n\tNo gene locus found for $equivalent_blast_hits[$qq] with $geneID in $fasta_header\n\n" if ($chr eq '' and $geneID ne '');

Sebastien Moretti's avatar
Sebastien Moretti committed
378
                    # Get Chr seq
379
380
                    if ( $chr ne '' ){
                        download_seq($cache, $amont, $aval, $chr);
Sebastien Moretti's avatar
Sebastien Moretti committed
381
                        $intronStep = 1;
382

Sebastien Moretti's avatar
Sebastien Moretti committed
383
                        unshift(@nt_GIs, "$chr:$amont-$aval");
384
                    }
385
386
387
388
389
                }
            }
        }


Sebastien Moretti's avatar
Sebastien Moretti committed
390
391
        # Align Nt seq with our protein query seq
        my ($POS, $BOJ) = runExonerate($cache, $date, $r, $original_seq[$r], $fasta_header, @nt_GIs);
392

Sebastien Moretti's avatar
Sebastien Moretti committed
393
        my ($POSresult, $bestNucleotide) = prepareResults4CDS($POS, $equivalent_blast_hits[$qq], $original_seq[$r], $fasta_header);
394
395
        $resultPOS .= $POSresult if ( $bestNucleotide ne '' && $POSresult ne '' && $resultPOS !~ /_G_$bestNucleotide/ );

396
397
        @intronStatus = ($POS->{0}, $equivalent_blast_hits[$qq], '')              if ( $intronStep==1 && exists($POS->{0}) && $POS->{0} ne '' && $POS->{0} =~ /\:/ );
        @intronStatus = ($POS->{0}, $equivalent_blast_hits[$qq], $bestNucleotide) if ( $intronStep==0 && ($bestNucleotide =~ /^N[CTWZG]_/ || $bestNucleotide =~ /^AC_/) );
398
399
400
401

        my ($intronLess, $nameLess) = ('', '');
        $intronLess = $POS->{0}       if ( %$BOJ eq 0 && exists($POS->{0}) && $POS->{0} ne '' );
        $nameLess   = $bestNucleotide if ( %$BOJ eq 0 && $intronStep==0 && ($bestNucleotide =~ /^N[CTWZG]_/ || $bestNucleotide =~ /^AC_/) );
Sebastien Moretti's avatar
Sebastien Moretti committed
402
        my ($BOJresult, $bestGenomic) = prepareResults4BOJ($BOJ, $equivalent_blast_hits[$qq], $original_seq[$r], $fasta_header, $intronLess, $nameLess);
403
        $resultBOJ .= $BOJresult      if ( $bestGenomic ne '' && $BOJresult ne '' && $resultBOJ !~ /_G_$bestGenomic[ \-]/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
404
405
406
    }


407
    if ( $resultPOS eq '' ){
408
409
410
411
412
413
414
        if ( $failureStatus[0] ne '' ){
            buildFailureOutputFiles($r, $failureStatus[1],         $failureStatus[0],   '');
        }
        else {
            buildFailureOutputFiles($r, $equivalent_blast_hits[0], 'Alignment_failure', '');
        }
        next HIT_LINK;
Sebastien Moretti's avatar
Sebastien Moretti committed
415
    }
416
417
    elsif ( $resultBOJ eq '' ){
        if ( $failureStatus[0] eq 'PUI_unavailable' || $failureStatus[0] eq 'No_nt_link' ){
418
            buildFailureBOJOutputFile($failureStatus[1], $failureStatus[0], $fasta_header, $original_seq[$r]);
419
420
        }
        elsif ( $failureStatus[0] eq '' && $intronStatus[0] ne '' ){
421
            buildIntronlessBOJOutputFile($intronStatus[1], $intronStatus[0], $intronStatus[2], $fasta_header, $original_seq[$r]);
422
423
        }
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
424
            buildFailureBOJOutputFile($equivalent_blast_hits[0], 'Unavailable', $fasta_header, $original_seq[$r]);
425
        }
426
        createCDSOutputFile($resultPOS, $original_seq[$r], $fasta_header);
Sebastien Moretti's avatar
Sebastien Moretti committed
427
428
    }
    else {
429
        createCDSOutputFile($resultPOS, $original_seq[$r], $fasta_header);
Sebastien Moretti's avatar
Sebastien Moretti committed
430
        createBOJOutputFile($resultBOJ, $original_seq[$r]);
Sebastien Moretti's avatar
Sebastien Moretti committed
431
432
433
434
435
436
    }


    undef $original_seq[$r];
    undef $original_names[$r];
}
437
close $LOG;
438
unlink "$date.log" if ( -z "$date.log" );
Sebastien Moretti's avatar
Sebastien Moretti committed
439
440


Sebastien Moretti's avatar
Sebastien Moretti committed
441
checkAndCleanStderrFiles("$cache/$date.ExonerateError") if ( -e "$cache/$date.ExonerateError" );
Sebastien Moretti's avatar
Sebastien Moretti committed
442
checkOutputFiles();
Sebastien Moretti's avatar
Sebastien Moretti committed
443
unlink("$msa") if ( $converted==1 );
444
unlink 'blast_result.txt';
445
my @tmpFiles = glob($cache."/${date}*");
Sebastien Moretti's avatar
Sebastien Moretti committed
446
if ( exists($tmpFiles[0]) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
447
448
    mkdir("$cache/$date.TMP");
    move("$_", "$cache/$date.TMP") while(<@tmpFiles>);
Sebastien Moretti's avatar
Sebastien Moretti committed
449
450
451
452
453
454
455
}


#Ouput Checker


#STDOUT Log output for our server
Sebastien Moretti's avatar
Sebastien Moretti committed
456
if ( -s "$originalMSA.cds" || -s "$originalMSA.out" ){
Sebastien Moretti's avatar
Sebastien Moretti committed
457
458
    print "\n\nTERMINATION STATUS: SUCCESS\n";
    print "OUTPUT RESULTS\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
459
460
461
462
463
464
465
466
467
468
    print "    #### File Type=        MSA Format= fasta_CDS Name= $originalMSA.cds\n" 
        if ( -s "$originalMSA.cds");
    print "    #### File Type=        MSA Format= fasta_CDS+Query Name= $originalMSA.cdsP\n" 
        if ( -s "$originalMSA.cdsP");
    print "    #### File Type=        MSA Format= colored_fasta_CDS Name= $originalMSA.cdsP.html\n" 
        if ( Views::Html("$originalMSA.cdsP") && -s "$originalMSA.cdsP.html");
    print "    #### File Type=        MSA Format= Rejected_seq Name= $originalMSA.out\n" 
        if ( -s "$originalMSA.out");
    print "    #### File Type=        MSA Format= fasta_Exon-Boundaries Name= $originalMSA.boj\n" 
        if ( -s "$originalMSA.boj" && -s "$originalMSA.cds" && $hideBOJ==0);
Sebastien Moretti's avatar
Sebastien Moretti committed
469
470
}
else{
Sebastien Moretti's avatar
Sebastien Moretti committed
471
    failure();
472
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
473
474
}

475
exit(0);
Sebastien Moretti's avatar
Sebastien Moretti committed
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519


=head1 NAME

ProtoGene.pl - Converts a peptidic alignment to a nucleotidic alignment through database search

=head1 SYNOPSIS

B<ProtoGene.pl> --msa=path_of_the_fasta_msa_file [options]

with these options available:

=over 4

=item I<--orgm>=targeted species in the database search

=item All_organisms [default], Bacteria, Viruses, Vertebrata, Eukaryota, Mammalia, Primates, Homo_sapiens, Gallus_gallus, Bos_taurus, Escherichia_coli, Arabidopsis_thaliana, Mus_musculus, Drosophila_Melanogaster, ...

=item I<--db>=targeted database in the database search

=item nr, pdb, swissprot, refseq_protein [default]

=item I<--local> to execute a local BLAST query with  --db=path_for_a_local_db_blast_formated

=item I<--template> to provide your own nucleotidic sequences following the cds file format

=item I<--revtrans> to reverse-translate sequences with no blast hit, in IUB (IUPAC) depiction code

=item They are removed from the alignement by default

=item I<--pep> to add the original peptide query beneath the back-translated sequence

=item I<--cache>=none, update, use, own_path_directory, old, empty

=item none: no cache usage, none temporary files are stored

=item update: use cache but update old files [default]

=item use: force use cache, whatever the age of files

=item own_path_directory: use my own directory, and its files

=item old: remove the old files in the cache directory

520
=item empty: empty completely the cache directory
Sebastien Moretti's avatar
Sebastien Moretti committed
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557

=item I<--version> to print version information

=item I<--help> to print this help message

=back

=head1 DESCRIPTION

PROTOGENE : Bona-Fide Back Translation of Protein Multiple Sequence Alignments
It converts a peptidic alignment to a nucleotidic alignment through database search.
Blast queries allow to get nucleotidic sequences from where peptidic sequences come from.
Exonerate aligns nucleotidic and peptidic sequences together.
PROTOGENE re-builds the original alignment with nucleotidic information it has gotten.

=head1 REQUIREMENT

=over 4

=item B<Perl 5.6 or better> is required !

=item Standard Perl modules B<lib>, B<strict>,  B<warnings>,  B<diagnostics>, B<Env> are required

=item and some other current ones : B<Getopt::Long>,  B<File::Which>,  B<File::Copy>, B<Mail::Send>

=item -

=item I<exonerate> from http://www.ebi.ac.uk/~guy/exonerate/

=item I<blast> from http://www.ncbi.nlm.nih.gov/BLAST/download.shtml or http://blast.wustl.edu/

=back

=head1 VERSION

=over 8

558
=item version 4.0.0
Sebastien Moretti's avatar
Sebastien Moretti committed
559

560
=item on Apr 02nd, 2011
Sebastien Moretti's avatar
Sebastien Moretti committed
561
562
563
564
565
566
567
568
569

=back

=head1 AUTHORS

=over 8

=item Sebastien MORETTI

Sebastien Moretti's avatar
Sebastien Moretti committed
570
=item moretti.sebastien [AT] gmail.com
Sebastien Moretti's avatar
Sebastien Moretti committed
571

572
=item Vital-IT computing center
Sebastien Moretti's avatar
Sebastien Moretti committed
573

574
=item Swiss Institute of Bioinformatics
Sebastien Moretti's avatar
Sebastien Moretti committed
575

576
=item Lausanne, Switzerland
Sebastien Moretti's avatar
Sebastien Moretti committed
577

578
=item http://www.vital-it.ch/
Sebastien Moretti's avatar
Sebastien Moretti committed
579
580
581
582
583
584
585
586

=back

=cut


######################  Management  ######################

587
# Failure declaration
Sebastien Moretti's avatar
Sebastien Moretti committed
588
sub failure {
589
590
591
    print {*STDERR} "\n\nFATAL\n\nTERMINATION STATUS: FAILURE\n\n";
    return;
}
Sebastien Moretti's avatar
Sebastien Moretti committed
592

Sebastien Moretti's avatar
Sebastien Moretti committed
593
# Check external programs presence in the PATH
594
sub checkExternalSoftware {
Sebastien Moretti's avatar
Sebastien Moretti committed
595
596
597
598
599
600
    for my $exe ( @_ ){
        if ( ! which($exe) ){
            failure();
            print {*STDERR} "\t'$exe' program is not reachable\n\tIt could not be in your PATH or not installed\n\n";
            exit(1);
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
601
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
602
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
603
604
}

605
606
# Check cache directory accessibility: Test and change rights, if required, of the cache directory
sub checkCacheAccessibility {
607
    my ($cacheDir) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
608

609
    if ( -d "$cacheDir/" && -w "$cacheDir/" && -x "$cacheDir/" && -r "$cacheDir/" ){
610
        # Good
Sebastien Moretti's avatar
Sebastien Moretti committed
611
    }
612
613
614
    elsif ( -d "$cacheDir/" ){
        if ( -o "$cacheDir/" ){
            chmod 0754, "$cacheDir/";
615
        }
616
        else{
617
618
619
            failure();
            print {*STDERR} "\tPermissions do NOT seem to be valid for the '$cacheDir' cache directory\n\n";
            exit(2);
620
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
621
622
    }
    else{
623
        mkdir "$cacheDir/"; # FIXME: mkdir is not recursive !
624
        if ( -o "$cacheDir/" ){
625
            chmod 0754, "$cacheDir/"; #0754 is in octal, so 754 or '0754' are not good !
626
627
        }
        else{
628
629
630
            failure();
            print {*STDERR} "\tPermissions do NOT seem to be valid for the '$cacheDir' cache directory\n\n";
            exit(2);
631
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
632
    }
633

634
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
635
636
}

637
# Cache Behavior Management
Sebastien Moretti's avatar
Sebastien Moretti committed
638
sub cacheManagement{
639
    my ($cache, $cacheDir, $inScript) = @_;
640

641
    if ( $cache eq 'empty' ){
642
        unlink <$cacheDir/*.fas>;
Sebastien Moretti's avatar
Sebastien Moretti committed
643
        unlink <$cacheDir/*.ExonerateError>;
644
645
        unlink("$cacheDir/webblast.log", "$cacheDir/web_tempo.result", "$cacheDir/debug.tempo");
        print "\n\tcache directory [$cacheDir] is empty\n\n";
646
        exit 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
647
    }
648
    elsif ( $cache eq 'old' ){
649
650
651
652
653
        unlink    grep { -M $_ > $cacheStorageTime }    glob($cacheDir.'/*.fas'); # Grep list only with $cacheStorageTime days old files

        if ( $inScript==0 ){
            print "\n\tcache directory [$cacheDir] is empty of its old files\n\n";
            exit 0;
654
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
655
    }
656
    elsif ( $cache eq 'update' || $cache eq 'use' ){
657
658
        # DEFAULT: Limit cached files to less than $cacheStorageTime days old ones for 'update'
        $cache = $cacheDir;
Sebastien Moretti's avatar
Sebastien Moretti committed
659
    }
660
    elsif ( $cache eq 'none' ){
661
        # Do not use cache
Sebastien Moretti's avatar
Sebastien Moretti committed
662
    }
663
664
665
666
    elsif ( -d $cache ){
        die "\n\tProtoGene cannot access to your directory\n\n"  if ( ! -x $cache );
        die "\n\tProtoGene cannot write into your directory\n\n" if ( ! -w $cache );
        die "\n\tProtoGene cannot read into your directory\n\n"  if ( ! -r $cache );
Sebastien Moretti's avatar
Sebastien Moretti committed
667
668
    }
    else {
669
670
671
        failure();
        print {*STDERR} "\n\tWrong cache argument, or your cache folder '$cacheDir' doesn't seem to exist\n\n";
        exit(3);
Sebastien Moretti's avatar
Sebastien Moretti committed
672
673
674
675
676
677
678
679
680
681
    }

    return($cache);
}

##########################################################


######################    Fasta    ######################

Sebastien Moretti's avatar
Sebastien Moretti committed
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
# Check input aln format
sub isFastaFile{
    my ($msa, $converted) = @_;

    if ( -s "$msa" ){ # Check is the file exists and is not empty
        # Is FASTA format ?
        my $firstLine = `head $msa`;
        return($msa, 0) if ( $firstLine =~ /^>/ );

        # Convert no fasta format to fasta with readseq
        my $reformat = which('readseq') || '';
        if ( $reformat ne '' ){
            system("$reformat -a -f8 $msa > $msa.007")==0 or do {failure(); print {*STDERR} "\n\tEnable to convert your file to fasta format\n\n";};
            $msa      .= '.007';
            $converted = 1;
697
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
698
699
    }
    else {
Sebastien Moretti's avatar
Sebastien Moretti committed
700
        failure();
701
702
        print {*STDERR} "\tThe file \'$msa\' doesn't not seem to be reachable\n\n";
        exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
703
704
    }

Sebastien Moretti's avatar
Sebastien Moretti committed
705
    return($msa, $converted);
Sebastien Moretti's avatar
Sebastien Moretti committed
706
707
708
709
}

#Prepare Fasta Header of the query name to incorporate our annotations
sub fastaHeaders4ProtoGene{
710
    my ($QueryName) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
711
712

    $QueryName =~ s/  +/ /g;
713
714
715
716
717
718
719
720
721
722
    my ($acc, $desc) = ($QueryName, $QueryName);
    $acc  =~ s/^ *([^ ]*) .*$/$1/;
    $acc  =~ s/\s+$//g;
    $acc  =  $acc.'_G_@@';

    $desc =~ s/^ *[^ ]* *(.*)$/$1/;
    $desc =~ s/^[\| \.]+//;
    $desc =~ s/[\|\. ]+$//;

    return($acc.$desc);
Sebastien Moretti's avatar
Sebastien Moretti committed
723
724
725
726
}

#Prepare Fasta Header to put it on the fasta header of the result outputs
sub getFastaHeaderAnnot{
727
    my ($bestTarget) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
728
729

    #Get annotation from nucleotide file header
Sebastien Moretti's avatar
Sebastien Moretti committed
730
731
    open(my $BEST, '<', "$cache/$bestTarget.fas");
    my $annot = <$BEST> || '';
732
    close $BEST;
Sebastien Moretti's avatar
Sebastien Moretti committed
733

Sebastien Moretti's avatar
Sebastien Moretti committed
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
    chomp($annot);
    $annot =~ s/^ *[^ ]* //;
    $annot =~ s/[\. ]*$//;
    $annot =~ s/^ *//;
    $annot =~ s/  +/ /g;

    return($annot);
}

#########################################################


###################### Translation ######################

#Direct Reverse translation
sub reverse_trad{
750
    my ($aa) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
751

752
753
754
755
756
757
    my %reverse_code = ('A' => 'GCN', 'F' => 'TTy', 'K' => 'AAr', 'P' => 'CCN', 'T' => 'ACN',
                        'C' => 'TGy', 'G' => 'GGN', 'L' => 'yTN', 'Q' => 'CAr', 'V' => 'GTN',
                        'D' => 'GAy', 'H' => 'CAy', 'M' => 'ATG', 'R' => 'mGN', 'W' => 'TGG',
                        'E' => 'GAr', 'I' => 'ATh', 'N' => 'AAy', 'S' => 'wsN', 'Y' => 'TAy',
                        'X' => 'NNN', '-' => '---', '*' => 'Trr',
                       ); # '*' is for stop codon
Sebastien Moretti's avatar
Sebastien Moretti committed
758

759
    my $triplet = 'nnn';
760
    while( my ($x, $y) =each(%reverse_code) ){
761
762
763
764
        if ( uc($aa) eq $x ){
            $triplet = $y;
            last;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
765
766
767
768
769
770
771
772
773
    }

    return($triplet);
}

#########################################################


####################### webblast #######################
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
sub display_blast_param {
    print "
             Program :                   BLASTp
             Evalue threshold :          $blast_param->{'evalue'}
             Matrix :                    $blast_param->{'matrix'}
             Filter :                    $blast_param->{'filter'}
             Equivalent best hits used : $blast_param->{'nbesthits'}

             Database:                   NCBI $blast_param->{'db1'}
             Blast_identity_threshold :  $blast_param->{'identity1'}
             Cover threshold :           $blast_param->{'cover1'}
          if no match
             Database:                   NCBI $blast_param->{'db2'}
             Blast_identity_threshold :   $blast_param->{'identity2'}
             Cover threshold :            $blast_param->{'cover2'}
\n";
    return;
}

sub run_BLAST{
    my ($infile, $whatBlast, $db, $blast_exe, $outfile) = @_;

    my $blast_at = $whatBlast==2  ? '-gigablast=yes'
                 : $whatBlast==1  ? "-blast_dir=$blast_exe"
                 :                  '';

    # BLAST run 1
    system("$webblast_exe -program=blastp -database=$db $blast_at -infile=$infile -matrix=$blast_param->{'matrix'} -evalue=$blast_param->{'evalue'} -method=geneid -filter=$blast_param->{'filter'} -organism=$species -identity=$blast_param->{'identity1'} -cover=$blast_param->{'cover1'} -hits=$blast_param->{'nbesthits'} -quiet=on -outfile=$outfile");


    if ( -z "$outfile" ){
    # BLAST run 2 if no hit with 100% identity and coverage: refseq100 -> nr95  OR  nr100 -> nr95
        print "run BLAST.. against $blast_param->{'db2'} because there was no acceptable hit against $blast_param->{'db1'}"
807
            if ( $whatBlast != 1 && $db ne 'nr' );
808
        print "run BLAST.. again, with decreased thresholds, because there was no acceptable hit against $db"
809
            if ( $whatBlast == 1 || $db eq 'nr' );
810
811

        system("$webblast_exe -program=blastp -database=$blast_param->{'db2'} $blast_at -infile=$infile -matrix=$blast_param->{'matrix'} -evalue=$blast_param->{'evalue'} -method=geneid -filter=$blast_param->{'filter'} -organism=$species -identity=$blast_param->{'identity2'} -cover=$blast_param->{'cover2'} -hits=$blast_param->{'nbesthits'} -quiet=on -outfile=$outfile");
Sebastien Moretti's avatar
Sebastien Moretti committed
812
813
    }

814
815
816
817
818
819
820
    move('webblast.log',     "$cache"); #Move temporary webblast.pl files
    move('web_tempo.result', "$cache");
    move('debug.tempo',      "$cache");

    unlink("$cache/webblast.log")     if ( -z "$cache/webblast.log" );
    unlink("$cache/web_tempo.result") if ( -z "$cache/web_tempo.result" );
    unlink("$cache/debug.tempo")      if ( -z "$cache/debug.tempo" );
Sebastien Moretti's avatar
Sebastien Moretti committed
821

822
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
823
824
825
826
827
828
829
}


########################################################


##################### NCBI requests #####################
830
831
832
sub fetch {
    my ($url) = @_;

833
    XML:
834
835
    for (my $tries=0; $tries <20; $tries++ ){
        my $content = get($url);
836
837
838
839
840
        print "[$content]\n\n" if ($debug);
        if ( defined $content ){
            next XML if ( $content =~ /<ERROR>/i );
            return $content;
        }
841
    }
842
843
    print {*STDERR} "Problem with NCBI eutils, please try again later\n";
    exit(40);
844
}
845

846
847
848
sub fetch_fasta {
    my ($url, $outfile) = @_;

849
    return 1 if ( -s "$outfile" && $cache eq 'update' && -M "$outfile" <= $cacheStorageTime );
850

851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
    GET_FASTA:
    for (my $tries=0; $tries <20; $tries++ ){
        my $HTTP_status = getstore($url, $outfile);
        if ( is_error($HTTP_status) ){
            next GET_FASTA;
        }
        else {
            open(my $FASTA, '<', "$outfile");
            my $counter = 0;
            my $lines   = 0;
            while(<$FASTA>){
                $lines++              if ( $_ !~ /^>/ );
                $counter = $counter+2 if ( $counter==1 && $_ !~ /^\w/ && $lines==1 && $_ !~ /^>/ );
                $counter++            if ( $_ =~ /^>/ );
                $counter = $counter+2 if ( $_ !~ /^>/ && ($_ =~ /Error:/ || $_ =~ /[<>]/) );
            }
            close $FASTA;
868

869
870
871
872
873
874
            return 1 if ( $counter == 1 );
        }
    }
    unlink("$outfile");
    print {*STDERR} "Problem with NCBI eutils, please try again later.\n";
    exit(45);
875
876
}

Sebastien Moretti's avatar
Sebastien Moretti committed
877
878
879

#Prot ACC -> PUID
sub blastPAcc2PGI{
880
    my ($blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
881

882
    my $protGI = '';
883
    #FIXME: should be ${blastHit}[pacc] but something is broken at NCBI
884
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=protein&term=$blastHit&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");
885
886
    if ( $content =~ /<Id>(\d+)<\/Id>/ ){
        $protGI = $1;
Sebastien Moretti's avatar
Sebastien Moretti committed
887
888
889
890
891
892
893
    }

    return($protGI);
}

#PUID -> nt GIs
sub protGI2NTGIs{
894
    my ($protGI, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
895

896
897
    my $ntGIs  = '';
    my $geneID = '';
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=protein&db=nuccore,nucleotide,gene&id=$protGI&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");

    my @xml  = split("\n", $content);
    my $flag = 0;
    for my $line (@xml){
        if (    $line =~ /<LinkName>protein_nuc[a-z]+<\/LinkName>/ ){ #for nuccleotide and nuccore
            $flag = 1;
        }
        elsif ( $line =~ /<LinkName>protein_gene<\/LinkName>/ ){
            $flag  = 2;
        }
        elsif ( $flag==1 && $line =~ /^.*<Id>(\d+)<\/Id>.*$/ ){
            my $match = $1;
            $ntGIs  .= "$match,$match,";
        }
        elsif ( $flag==2 && $line =~ /^.*<Id>(\d+)<\/Id>.*$/ ){
            my $match = $1;
            $geneID .= "$match,$match,";
916
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
917
918
    }

919
    #Remove redundancy if any
920
    chop $ntGIs; #Remove last ','
921
922
923
924
925
926
    my %hash_NT = split(',', $ntGIs);
    $ntGIs      = join(',',  keys(%hash_NT) );
    chop $geneID;
    my %hash_GE = split(',', $geneID);
    $geneID     = join(',',  keys(%hash_GE) );

927
    return($ntGIs, $geneID);
Sebastien Moretti's avatar
Sebastien Moretti committed
928
929
930
}

sub geneID2Chr{
931
    my ($geneID, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
932

933
934
    my $chr   = '';
    my ($amont, $aval) = ('', '');
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
#   my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=gene&id=$geneID&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gene&id=$geneID&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");

    my @xml  = split("\n", $content);
    my $flag = 0;
    GENE_INFO:
    for my $line (@xml){
        if ( $flag==0    && $line =~ /<Item Name="GenomicInfoType" Type="Structure">/ ){
            $flag = 1;
        }
        elsif ( $flag==1 && $line =~ /<Item Name="ChrAccVer" Type="String">([^<]+?)\.?\d*<\/Item>/ ){
            $chr   = $1;
            $flag  = 3;
        }
        elsif ( $flag==3 && $line =~ /<Item Name="ChrStart" Type="Integer">(\d+)<\/Item>/ ){
            $amont = $1;
            $flag  = 4;
        }
        elsif ( $flag==4 && $line =~ /<Item Name="ChrStop" Type="Integer">(\d+)<\/Item>/ ){
            $aval  = $1;
            last GENE_INFO;
956
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
957
    }
958

959
960
    if ( $amont eq '' || $aval eq '' ){
        $amont = '';
961
        $aval  = '';
Sebastien Moretti's avatar
Sebastien Moretti committed
962
    }
963
964
    elsif ( $amont ne '' && $aval ne '' ){
        if ( $amont > $aval ){
965
966
            $amont = $amont + 5000;
            $aval  = $aval  - 5000;
967
968
        }
        else{
969
970
            $amont = $amont - 5000;
            $aval  = $aval  + 5000;
971
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
972
    }
973
    $amont = 1 if ( $amont ne '' && $amont <= 0 );
974
    $aval  = 1 if ( $aval  ne '' && $aval  <= 0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
975

976
    return($chr, $amont, $aval);
Sebastien Moretti's avatar
Sebastien Moretti committed
977
978
979
}

sub downloadSeqFromGIs{
980
    my ($cache, $amont, $aval, @acc) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
981

982
    GET_SEQ:
Sebastien Moretti's avatar
Sebastien Moretti committed
983
    for(my $a=0; $a<=$#acc; $a++){
984
        if ( $amont =~ /^\d+$/ && $aval =~ /^\d+$/ ){
985
            fetch_fasta("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nucleotide&id=$acc[$a]&rettype=fasta&retmode=text&from=$amont&to=$aval&tool=ProtoGene&email=smoretti\@unil.ch", "$cache/$acc[$a]-$amont.fas") ;
986
987
        }
        else{
988
            fetch_fasta("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nucleotide&id=$acc[$a]&rettype=fasta&retmode=text&tool=ProtoGene&email=smoretti\@unil.ch'", "$cache/$acc[$a].fas");
989
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
990
    }
991
992

    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
993
994
}

Sebastien Moretti's avatar
Sebastien Moretti committed
995
sub download_seq{
996
    my ($cache, $amont, $aval, @acc) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
997

998
999
1000
    my $cp   = 0;
    my $from = $amont;
    my $to   = $aval;